Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Preferred Deals in General Environments
Authors: Yuan Deng, Sébastien Lahaie, Vahab Mirrokni
IJCAI 2019 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate our algorithm using auction data from a major advertising exchange and our empirical results show that the algorithm achieves around 95% of the optimal revenue. |
| Researcher Affiliation | Collaboration | Yuan Deng1 , S ebastien Lahaie2 and Vahab Mirrokni2 1Duke University 2Google Research EMAIL, EMAIL |
| Pseudocode | Yes | Algorithm 1: AAG Framework |
| Open Source Code | No | The paper refers to using the Glop linear programming solver and provides a link to its documentation, but does not provide access to the authors' own implementation code. |
| Open Datasets | No | We use data collected from the Google Ad Exchange (Ad X) over a period of one day in summer 2018. |
| Dataset Splits | No | The paper describes the data collection and instance creation (100K auctions) and how experiments are repeated (50 times) with varying budget ratios, but it does not specify explicit train/validation/test dataset splits. |
| Hardware Specification | No | Each run of the experiment takes roughly 30 seconds on a single CPU. |
| Software Dependencies | No | The paper mentions 'Python 2.7' and 'Glop linear programming solver' but does not provide specific version numbers for the Glop solver or other required libraries. |
| Experiment Setup | Yes | We run our experiment on 5 high-volume inventory units for the day in question. ... we discretize the bids to cents and only consider the top 50 most frequent buyer-bid pairs. ... take the first 100K auctions in which at least two of the top 50 buyer-bid pairs appear in the auction to form our instances. ... We conduct experiments parametrized by a budget ratio r [0.1, 1.5]. ... For a fixed setting of r, we repeat the experiment 50 times and generate the budget as follows: for each run, (1) compute the contribution si to the social welfare for each buyer i, by summing buyer i s bids over all auctions that it wins; (2) set buyer i s budget to a value uniformly drawn from [0, 2 si r], so that the mean of the generated budget is si r, proportional to the buyer s social welfare contribution. |